| name | kachilu-browser |
| description | Browser automation CLI for AI agents. Use when the user needs to interact with websites, including navigating pages, filling forms, clicking buttons, taking screenshots, extracting data, testing web apps, logging in to sites, posting to social apps, handling CAPTCHA workflows, or automating browser actions. Load the core guide to choose between host-provided sessions, MCP tools, and CLI commands. |
| allowed-tools | Bash(npx kachilu-browser:*), Bash(kachilu-browser:*) |
| hidden | true |
kachilu-browser
Browser automation CLI for AI agents. Chrome/Chromium via CDP with
accessibility-tree snapshots and compact @eN element refs.
Install: npm i -g kachilu-browser && kachilu-browser install
Start Here
This file is a discovery stub, not the usage guide. Before running any
kachilu-browser command, load the actual workflow content from the CLI:
kachilu-browser skills get core
kachilu-browser skills get core --full
The CLI serves skill content that matches the installed version, so instructions
do not drift when the package is upgraded. The core guide contains the Kachilu
human-like interaction defaults, control-plane routing rules, rich composer safeguards,
CAPTCHA workflow, and the command reference.
If the host already provides a pre-attached browser session, CDP endpoint,
session ref file, or explicit kachilu-browser --session ... --cdp ... command,
use that host-provided session through the CLI. Do not switch to MCP
prepare_workspace or auto-connect unless the core guide says that no
host-provided session is available.
For MCP hosts such as OpenClaw, keep using the same prepared workspace session
across related browser work in one user request, even when moving between sites
such as LinkedIn and X. The site hint is for routing only. Do not close the
workspace between those steps unless the user explicitly wants cleanup. If the
daemon still exists but the browser connection is gone, treat that workspace as
stale and reconnect instead of reusing it.
Critical Fallback
If an element-targeted click fails but the visible cursor is already on the
intended button or control, take a screenshot/capture and confirm the cursor is
on target. Then click at the current cursor position:
kachilu-browser mouse down left
kachilu-browser mouse up left
kachilu-browser snapshot -i
Use this after visual confirmation instead of repeating the same failing ref or
locator click.